Backslashes in regex_replace

[kustodian]

ISSUE TYPE

• Bug?

COMPONENT NAME

Core

ANSIBLE VERSION

ansible 2.5.0 (devel 1ee511f82c) last updated 2017/11/22 20:37:38 (GMT +200)

SUMMARY

While I was upgrading from Ansible 2.3 to 2.4 I came across some changes in how backslashes are used to escape characters in regex_replace filter, so I made a little test playbook to show all the correct ways you can use backslashes. I told @abadger that I will make an issue for this, so here it is.

STEPS TO REPRODUCE

Here is a test playbook:

# fun_with_backslashes.yml
---
- hosts: localhost
  connection: local
  gather_facts: no
  vars:
    full_name: 'John Smith'
  tasks:
    - name: Backslashes msg in single quotes
      debug:
        msg: '{{ full_name | regex_replace("^([\S]+).*", "\1") }}'

    - name: Backslashes msg in double quotes
      debug:
        msg: "{{ full_name | regex_replace('^([\\S]+).*', '\\1') }}"

    - name: Backslashes with var using single quotes
      debug:
        var: full_name | regex_replace('^([\S]+).*', '\1')

    - name: Backslashes with var using double quotes
      debug:
        var: full_name | regex_replace("^([\S]+).*", "\1")

    - name: Backslashes in a template
      template:
        src: backslashes.j2
        dest: /tmp/backslashes
      check_mode: yes

    - name: Lookup template
      debug:
        msg: '{{ lookup("template", "backslashes.j2") }}'

and the template:

{# backslashes.j2 #}
{{ full_name | regex_replace("^([\\S]+).*", "\\1") }}
{{ full_name | regex_replace("^([\S]+).*", "\\1") }}
{% if full_name | regex_replace("^([\\S]+).*", "\\1") == 'John' %}
John
{% endif %}
{% if full_name | regex_replace("^([\S]+).*", "\\1") == 'John' %}
John
{% endif %}

These are all the combination I could think of that work correctly. The output of running this playbook (with -D to be able to see results of the template module):

$ ansible-playbook fun_with_backslashes.yml -D

PLAY [localhost] **********************************************************************************************************************************************************************************************

TASK [Backslashes msg in single quotes] ***********************************************************************************************************************************************************************
ok: [localhost] => {
    "msg": "John"
}

TASK [Backslashes msg in double quotes] ***********************************************************************************************************************************************************************
ok: [localhost] => {
    "msg": "John"
}

TASK [Backslashes with var using single quotes] ***************************************************************************************************************************************************************
ok: [localhost] => {
    "full_name | regex_replace('^([\\S]+).*', '\\1')": "John"
}

TASK [Backslashes with var using double quotes] ***************************************************************************************************************************************************************
ok: [localhost] => {
    "full_name | regex_replace(\"^([\\S]+).*\", \"\\1\")": "John"
}

TASK [Backslashes in a template] ******************************************************************************************************************************************************************************
--- before
+++ after: /tmp/tmpy3Q4hm/backslashes.j2
@@ -0,0 +1,4 @@
+John
+John
+John
+John

changed: [localhost]

TASK [Lookup template] ****************************************************************************************************************************************************************************************
ok: [localhost] => {
    "msg": "John\nJohn\nJohn\nJohn\n"
}

EXPECTED RESULTS

Results mostly look as expected:

• Single backslash:
  • in YAML between single quotes
  • in YAML without quotes (doesn't matter if you use single or double quotes inside regex_replace)
• Double backslash:
  • in YAML between double quotes
  • in templates

ACTUAL RESULTS

Only strange thing is that inside templates you can use both double and single quotes inside a regex, but you have to use double quotes for back reference, which means both of these lines work in templates

{{ full_name | regex_replace("^([\\S]+).*", "\\1") }}
{{ full_name | regex_replace("^([\S]+).*", "\\1") }}

In my opinion the second line feels like a bug and it should be consistent.

[sivel]

This change was expected, and indicated in the 2.4 porting guide https://docs.ansible.com/ansible/latest/porting_guides/porting_guide_2.4.html#template-lookup-plugin-escaping-strings